bpo-39942:Fix TypeVar fails when missing __name__
#19616
Conversation
There was a problem hiding this comment.
Thanks for opening a pull request @hongweipeng, I tested the changes and it works well, I'm not sure if defaulting to __main__ is the correct behavior thought. If I'm correct both Enums and dataclasses had this issue and they are now just unpickable when __name__ cannot be determined.
| if def_mod != 'typing': | ||
| self.__module__ = def_mod | ||
| try: | ||
| def_mod = sys._getframe(1).f_globals.get('__name__', '__main__') # for pickling |
There was a problem hiding this comment.
I don't think it's safe to assume __name__ is __main__ when it's not defined here. I think other classes like enums make the class unpickable when __name__ cannot be determined.
There was a problem hiding this comment.
I think '__main__' is fine. By the way a similar code is used by TypedDict and NamedTuple, so you can copy it here. Or maybe it would even make sense to factor this out in an internal helper function doing this:
try:
module = sys._getframe(1).f_globals.get('__name__', '__main__')
except (AttributeError, ValueError):
module = None| @@ -221,6 +221,13 @@ def test_bound_errors(self): | |||
| with self.assertRaises(TypeError): | |||
| TypeVar('X', str, float, bound=Employee) | |||
|
|
|||
| def test_missing__name__(self): | |||
There was a problem hiding this comment.
def test_missing__name__(self):
# See bpo-39942
code = ("import typing\n"
"T = typing.TypeVar('T')\n"
)
exec(code, {})would be more like the other tests
|
This change will also require a NEWS entry (https://devguide.python.org/committing/#what-s-new-and-news-entries), you can add one by using https://blurb-it.herokuapp.com/ |
|
Actually, it probably makes sense to backport this to 3.7 and 3.8 since this is technically a bug. Let me try if this still can be done automatically. |
|
Thanks @hongweipeng for the PR, and @ilevkivskyi for merging it 🌮🎉.. I'm working now to backport this PR to: 3.7. |
|
Thanks @hongweipeng for the PR, and @ilevkivskyi for merging it 🌮🎉.. I'm working now to backport this PR to: 3.8. |
…19616) https://bugs.python.org/issue39942 (cherry picked from commit a25a04f) Co-authored-by: HongWeipeng <hongweichen8888@sina.com>
|
GH-19626 is a backport of this pull request to the 3.7 branch. |
…19616) https://bugs.python.org/issue39942 (cherry picked from commit a25a04f) Co-authored-by: HongWeipeng <hongweichen8888@sina.com>
|
GH-19627 is a backport of this pull request to the 3.8 branch. |
https://bugs.python.org/issue39942 (cherry picked from commit a25a04f) Co-authored-by: HongWeipeng <hongweichen8888@sina.com>
https://bugs.python.org/issue39942 (cherry picked from commit a25a04f) Co-authored-by: HongWeipeng <hongweichen8888@sina.com>
https://bugs.python.org/issue39942